基于非均匀测量矩阵的超分辨率全向图像重建  

Super-resolution omnidirectional image reconstruction based on non-uniform measurement matrix

作　　者：娄静涛[1] 谭树人[1] 李永乐[1] 王炜[1] 张茂军[1] 

机构地区：[1]国防科技大学信息系统与管理学院,湖南长沙410073

出　　处：《光电子．激光》2013年第12期2421-2429,共9页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金(61175006;61175015;61275016;61271438)资助项目

摘　　要：为解决折反射全向成像空间分辨率低和分布不均匀的问题,将压缩感知(CS)理论引入折反射全向成像系统。基于单位球面模型分析了折反射全向成像系统的空间分辨率分布规律,根据重构信号的均方误差(MSE)与稀疏度、测量次数的关系,设计了基于全向图像分辨率的非均匀测量矩阵;通过设计的测量矩阵,将较多的传感资源分配给全向图像内环,而将较少的传感资源分配给外环,从而对经过镜面反射的场景进行采集,得到观测场景的非均匀压缩采样;最后通过线性Bregman迭代算法重构出分辨率均匀的高分辨率图像。实验结果表明,本文方法得到的图像空间分辨率更高且分布更为均匀,有效改善了全向成像的分辨率问题。To solve the problem of low and non-uniform resolution in catadioptric omnidirectional imaging, the theory of compressive sensing （CS） is applied to research the catadioptric omnidirectional imaging systems. The resolution distribution of eatadioptric omnidirectional imaging systems is analyzed based on the unit sphere model. According to the relation between mean square error （MSE） and sparsity,measurement number,a non-uniform measurement matrix, which is based on the distribution of the imaging systemPs resolution, is designed in this paper. Based on the designed measurement matrix,it allocates more sensing resources to inner parts but fewer to outer parts of the catadioptric omnidirectional image. This scheme takes the resolution of omnidirectional image into account, which is increased from the center to the periphery. The non-uniform compressed samples of the observed scene are captured from the ray light which is reflected from the mirror. The linear Bregrnan iteration is employed as the reconstruction algorithm to obtain the high and uniform resolution image. The algorithm is tested on synthetic and realistic images. Experimental results show that the proposed method is feasible and effective. The recovered image has higher and much more uniform resolution than that reconstructed by traditional method.

关 键 词：折反射全向成像 压缩感知(CS) 分辨率问题 超分辨率重建 测量矩阵 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]